Build a 13-week rolling cash flow forecast from bank exports, AR aging, AP bills, payroll, and recurring commitments -- the report that tells a small business whether it survives the quarter. Flags the crunch weeks early enough to act.
SkillScore74Maintained
Automatic
◆Finance & Accounting pack
Last verified 8/24/2026
What it changes
Passed 17 of 23 test cases with the skill, and 12 of 23 without it.
Without skill52%
With skill74%
Checks it passed that the model alone missed
The action pairs with cowork-invoice-chaser.
Week 3 is designated as the headline crunch week.
Week 4 is designated as the primary headline crunch week.
and 3 more in the full report
6 fixed11 passed either way5 still failing1 broke
Pass-rate lift
+22 pp
Tokens
+13%
Agent turns
0%
Response time
-28%
Model gemini-3.6-flashMeasured 8/24/2026v117 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 6
Prompt
In our projected balance, week 4 ends at $5,000 (below our $10,000 comfort level) and week 9 ends at -$25,000. Which week is designated as the primary headline crunch?
What was checked, without and with the skill
Checked
Without skill
With skill
Week 4 is designated as the primary headline crunch week.
Missed
Passed
case-19 · gemini-3.6-flash · 8/24/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md27 linessingle file
Cash Flow Forecaster
Profitable businesses die of cash starvation; the 13-week forecast is the instrument that prevents it. Input: bank/transaction exports (.csv/.xlsx), open invoices (AR), upcoming bills (AP), payroll schedule, and known recurring items. Output: a week-by-week cash position with the crunch points circled.
Workflow
Establish the baseline. Current cash across accounts (from the most recent export -- state the as-of date prominently). Reconstruct 8-12 weeks of history to learn the rhythm: payroll cadence, rent day, typical weekly card/vendor spend, revenue deposit patterns.
Schedule the knowns. Inflows: open AR placed in the week each invoice is likely paid -- due date plus that customer's historical lateness, not the due date printed on the invoice. Outflows: payroll (with tax deposit timing), rent, loan payments, insurance, subscriptions, credit card due dates, quarterly estimated taxes -- the ambush everyone forgets.
Model the unknowns. Recurring-but-variable spend from historical averages, labeled ESTIMATE. New revenue only if the user provides expected deals -- never invent pipeline.
Build the forecast.cash-flow-13wk.csv + summary: weekly beginning cash, inflows, outflows, ending cash. Flag every week ending below the user's minimum comfort level (ask; default one payroll cycle) as CRUNCH, and the first crunch week is the headline.
Scenario the levers. For each crunch: what closes the gap -- which specific AR to chase this week (pairs with cowork-invoice-chaser), which AP can slide two weeks without damage, where the line of credit covers, what the owner draw pause buys. Concrete moves with amounts, not advice-shaped sentences.
Rules
Payment behavior beats due dates: a net-30 customer who pays in 55 days forecasts at 55.
Never fabricate inflows. Hope is not a week-9 deposit; unconfirmed revenue stays out or sits in a clearly separated optimistic scenario.
Estimates are labeled and totaled separately so the user can see how much of the forecast is soft.
Weekly granularity, not monthly -- a month that nets positive can contain a week that bounces payroll.
The as-of date and data gaps are stated up front; a forecast built on a two-week-old balance says so.
Re-run weekly: roll the window, compare actuals to last week's forecast, and report the misses -- forecast accuracy is a metric.
Quick Commands
"Forecast from [files]" -- full workflow
"When's my next crunch?" -- the headline week and its gap
"What if [customer] pays late?" -- single-scenario rerun
"Roll the forecast" -- weekly update against actuals